Dear Tom,
I guess with "modulated" you refer to the normalized, modulated mwc images. Then your data is already in MNI space and fits well with the AAL. Note that the AAL is based on a single subject, thus it might indeed be the case that atlas and images don't overlap perfectly.
Another issue is that AAL (as provided with the corresponding toolbox) has a resolution of 2x2x2 mm^3, while the output from segmentation is 1.5x1.5x1.5 mm^3 as far as I remember. Thus you would go with a Coregister: Reslice step, for image defining space you would load one of the mwc files, for images to reslice the ROI_MNI_V4.nii, and for reslice options / interpolation go with "Nearest neighbour". The resliced AAL file could then be applied to all your data (maybe you've already done so anyway). If you encounter issues with GM voxels not covered by AAL labels you could apply some smoothing to the labels, but I'd suggest to just go with the resliced AAL files, as this is easy to replicate for other workgroups. In case AAL labels also cover non-GM voxels / voxels with low GM volume you might want to threshold individual mwc files, so that only voxels with a certain intensity like .1 or .2 are really interpreted as GM and considered for volume calculation, with the other voxels being discarded. This is going to depend on your definitions of GM volume.
In general, as the AAL is just based on a single subject, you might want to turn to probabilistic brain atlases based on several subjects like the Hammers-mith n30r83 http://biomedic.doc.ic.ac.uk/brain-development/index.php?n=Main.AdultMaxProb or the LPBA40 http://www.loni.usc.edu/atlases/Atlas_Detail.php?atlas_id=12 . SPM12 also has some labels stored in tpm\labels_Neuromorphometrics.nii, for details see spm_templates.man.
Best,
Helmut
|